Book contents
- Frontmatter
- Contents
- Preface
- Acronyms and abbreviations
- Principal symbols
- 1 Introduction
- 2 The governing systems of equations
- 3 Numerical solutions to the equations
- 4 Physical-process parameterizations
- 5 Modeling surface processes
- 6 Model initialization
- 7 Ensemble methods
- 8 Predictability
- 9 Verification methods
- 10 Experimental design in model-based research
- 11 Techniques for analyzing model output
- 12 Operational numerical weather prediction
- 13 Statistical post processing of model output
- 14 Coupled special-applications models
- 15 Computational fluid-dynamics models
- 16 Climate modeling and downscaling
- Appendix: Suggested code structure and experiments for a simple shallow-fluid model
- References
- Index
9 - Verification methods
Published online by Cambridge University Press: 05 June 2012
- Frontmatter
- Contents
- Preface
- Acronyms and abbreviations
- Principal symbols
- 1 Introduction
- 2 The governing systems of equations
- 3 Numerical solutions to the equations
- 4 Physical-process parameterizations
- 5 Modeling surface processes
- 6 Model initialization
- 7 Ensemble methods
- 8 Predictability
- 9 Verification methods
- 10 Experimental design in model-based research
- 11 Techniques for analyzing model output
- 12 Operational numerical weather prediction
- 13 Statistical post processing of model output
- 14 Coupled special-applications models
- 15 Computational fluid-dynamics models
- 16 Climate modeling and downscaling
- Appendix: Suggested code structure and experiments for a simple shallow-fluid model
- References
- Index
Summary
Background
What is verification?
Forecast verification involves evaluating the quality of forecasts. Various methods exist to accomplish this. In all cases, the process entails comparing model-predicted variables with observations of those variables. The term validation is sometimes used instead of verification, but the intended meaning is the same. That said, the root word “valid” may imply to some that a forecast can either be valid, or invalid, whereas obviously there is a continuous scale that measures forecast quality. Thus, the term verification is preferable to many, and will be employed here. Special verification measures that are most applicable to ensemble predictions have been discussed in Chapter 7. There is an extensive body of literature on the subject of model verification, and students and researchers should read beyond the summary material in this chapter to ensure that they understand underlying statistical concepts and that they use the verification metrics that are most appropriate for their needs.
Reasons for verifying model simulations and forecasts
There are multiple motivations for evaluating the quality of model forecasts or simulations.
Most models are under continuous development, and the only way modelers can know if routine system changes, upgrades, or bug fixes improve the forecast or simulation quality is to objectively and quantitatively calculate error statistics.
For physical-process studies, where the model is used as a surrogate for the real atmosphere, the model solution must be objectively verified using observations, and if the observations and model solution correspond well where the observations are available, there is some confidence that one can believe the model where there are no observations.
[…]
- Type
- Chapter
- Information
- Numerical Weather and Climate Prediction , pp. 294 - 320Publisher: Cambridge University PressPrint publication year: 2010