On Right Solutions and Right Equations. The Relationship Between Verification and Validation of Climate Simulations

Beisbart, Claus (12 September 2019). On Right Solutions and Right Equations. The Relationship Between Verification and Validation of Climate Simulations (Unpublished). In: EPSA19. Université de Genève. 11.-14.9.2019.

Climate simulations need to be tested using the methods of verification and validation (V&V). Here, verification is supposed to be about „solving the equations right“, whereas validation is expected to ensure „solving the right equations“ (e.g. Roache). While some authors, e.g., Oberkampf, draw a clear line between verification and validation and demand that the former precede the latter, some philosophers, e.g., Winsberg, have suggested that the distinction has been over-exaggerated in climate science and similar fields. The aim of this talk is to clarify the relationship between verification and validation of climate simulations.
To this purpose, I distinguish between the simulation and its conceptual model and argue that both may be the proper object of validation. Using the so-called Sargent cycle, it is easy to see that verification and validation of the simulation or the conceptual model differ in terms of their primary goals and standards. Things become more complicated only because some of the methods can be placed into service of others. For instance, scientists may in principle validate a simulation by validating the conceptual model independently and by verifying the simulation with respect to it. But this is unrealistic in climate science, because the conceptual model cannot be validated independently. What is common instead is a direct validation of the simulations using a comparison between simulation output and measured data. I argue that this strategy requires an inductive inference that is only rational if there is some prior confidence (i) in the conceptual model and (ii) in verification. Thus, verification is prior to the validation of simulation. Doubts that Winsberg and Lenhard have voiced about the separation of verification and validation in practice can be rejected. In particular, I show that the issue of life cycles of models is a red herring in the debate.

Interest & Impact

Downloads

0 since deposited on 28 Apr 2020
0 in the past 12 months

Citations

5 Citations in Web of Science ®
6 Citations in Scopus

Search

in Google Scholar™

Services

Actions (login required)

Edit item

Actions (login required)

Edit item