Conceptual Overview
Numerical computing provides many of the mathematical and algorithmic tools used by scientific computing. It focuses on calculating numerical approximations to mathematical quantities and solving mathematical problems using finite-precision computers. Scientific computing includes these activities but also addresses how a real scientific problem is formulated, represented, simulated, evaluated, and interpreted. The two fields overlap extensively, and numerical computing is often a fundamental component of scientific computing.
Scientific Computing
Focus
The complete computational investigation of scientific and engineering problems.
Central Question
How can computational methods help us understand, investigate, simulate, or predict a real scientific system?
Typical Activities
Scientific problem formulation.
Mathematical and physical modelling.
Numerical algorithm selection.
Scientific software development.
Simulation and computational experimentation.
Scientific data processing and visualization.
Uncertainty analysis and sensitivity analysis.
Verification and validation.
Scientific interpretation and communication.
Outputs
Validated computational models.
Scientific simulations.
Numerical predictions.
Analyzed scientific datasets.
Computational experiments.
Scientific interpretations and evidence.
Numerical Computing
Focus
The numerical representation and computational solution of mathematical problems.
Central Question
How can a mathematical quantity or equation be calculated accurately and efficiently using finite-precision computer arithmetic?
Typical Activities
Solving linear and nonlinear equations.
Numerical differentiation and integration.
Matrix and vector computations.
Interpolation and approximation.
Numerical optimization.
Eigenvalue and eigenvector computation.
Numerical solution of ordinary and partial differential equations.
Analysis of numerical stability, accuracy, and convergence.
Outputs
Approximate numerical values.
Solutions to mathematical systems.
Computed matrices and vectors.
Numerical approximations of functions.
Estimates of integrals, derivatives, and roots.
Error and convergence measurements.
Comparison
| Dimension | Scientific Computing | Numerical Computing |
|---|---|---|
Scope | Broad interdisciplinary investigation of scientific problems. | More focused on numerical calculation and mathematical algorithms. |
Primary objective | Understand or investigate a scientific system using computational methods. | Calculate approximate solutions to mathematical problems. |
Problem origin | Often begins with an observation, scientific question, experiment, or engineering requirement. | Often begins with a mathematical expression, numerical dataset, or computational problem. |
Mathematical modelling | Usually includes selecting or constructing a model representing a real system. | May focus on solving a mathematical problem without developing the scientific model that produced it. |
Algorithms | Uses numerical algorithms alongside simulation, statistical, data-processing, and optimization techniques. | Emphasizes algorithms for numerical representation and mathematical calculation. |
Data | Can involve experimental observations, simulation outputs, scientific images, and large datasets. | Often works with numerical arrays, matrices, vectors, and mathematical input values. |
Validation | May require mathematical verification, experimental validation, and evaluation of scientific assumptions. | Frequently emphasizes numerical error, stability, convergence, and agreement with known mathematical solutions. |
Typical outcome | A scientifically interpreted computational result or model. | A numerical approximation or computed mathematical solution. |
Relationship
Explanation
Numerical computing is an important enabling area within scientific computing. A scientific simulation may depend on numerical integration, matrix operations, differential-equation solvers, or optimization algorithms. However, obtaining a numerical answer is only one part of a scientific investigation. Researchers must also determine whether the mathematical model represents the actual system and whether the computed result has scientific meaning.
Hierarchy
Numerical computing can be viewed as a specialized computational foundation used by scientific computing, although numerical computing also has applications outside scientific research.
Important Note
The terms are sometimes used differently across institutions, textbooks, and research communities. Their exact boundaries are not universally standardized.
Worked Examples
| Example | Problem | Method | Result | Interpretation |
|---|---|---|---|---|
Numerical computing example: solving a linear system | Solve the equations 2x + y = 5 and x + 3y = 6. | Represent the equations as a matrix system Ax = b and use Gaussian elimination or a numerical linear algebra solver. | The solution is x = 1.8 and y = 1.4. | The main task is obtaining a numerical solution to a mathematical system. |
Scientific computing example: modelling pollutant transport | Investigate how a pollutant spreads through a river over time. | Develop a transport model, define flow conditions and pollutant properties, discretize the governing equations, implement a numerical solver, and simulate different scenarios. | The numerical solver is one component of a larger scientific investigation involving environmental assumptions, measurements, model construction, validation, and interpretation. |
When They Overlap
When a scientific model is converted into equations and solved numerically.
When researchers evaluate the stability and convergence of a scientific simulation.
When matrix computations are used to process scientific measurements.
When numerical optimization estimates parameters in a physical or biological model.
When numerical differential-equation solvers calculate the evolution of a scientific system.
Common Misconceptions
Numerical computing is not separate from scientific computing; it is one of its major mathematical foundations.
Scientific computing is not simply a larger version of numerical computing because it also includes scientific reasoning, modelling, data interpretation, and validation.
A numerical solution can be mathematically accurate while the scientific model itself is inappropriate.
A scientific simulation can use statistical or data-driven methods that are not limited to traditional numerical equation solving.
Practical Learning
Solve a linear system manually and then using a numerical programming library.
Implement a numerical integration method and compare it with an analytical integral.
Use a numerical differential-equation solver to simulate a physical system.
Extend the differential-equation calculation into a scientific investigation by defining realistic assumptions, comparing results with observations, and analyzing limitations.
Identify which parts of a scientific simulation belong specifically to numerical computing.
Learning Objectives
Define scientific computing and numerical computing.
Explain the differences in their objectives and scope.
Understand how numerical methods support scientific investigations.
Recognize that numerical accuracy and scientific validity are different requirements.
Identify numerical computing tasks within a broader computational science workflow.
Assessment Questions
What is the primary difference between scientific computing and numerical computing?
Why is numerical computing considered an important foundation of scientific computing?
Can a numerically accurate solution still be scientifically misleading? Explain.
Which activities in a pollutant-transport simulation belong to numerical computing and which belong to scientific computing?
Why should researchers evaluate both mathematical accuracy and model validity?