Verification and Validation in Scientific Computing
The value of any scientific model is in it's ability to predict. The accuracy of the prediction of a scientific model developed for numerical solution has until recently been of low priority in plasma physics mainly due to the maturity of the computational model (only recently have plasma physics models reached a sufficient level of development that we can have confidence that they entail the necessary physics set). Other scientific disciplines have reached this point of maturity sooner, and are now engaged in a mature level of validation of code predictions, quantifying error and uncertainty in all parameters used and predicted by numerical simulations. This quantification of the predictive accuracy of code now informs research program priorities and helps guide decisions in program efforts and direction. A rich field of validation is now well established, epitomized in tomes such as Oberkampf's 2010 'Verification and Validation in Scientific Computing' Greenwald's `Verification and validation for magnetic fusion' and Terry's `Validation in Fusion Research: Towards Guidelines and Best Practices' (from which we draw considerable influence for this section of the course).
- Code verification
- Solution verification
- Model validation
- Validation metrics
- Predictive uncertainty