Two different questions get answered by two different kinds of evidence, and the words for them are routinely swapped. Verification asks whether the software solves the equations it claims to solve: does the answer stop changing as the grid is refined, is energy conserved, does the code reproduce a case with a known analytic solution? It is a question about mathematics and programming and can be settled without a single real fire. Validation asks whether those equations describe fire: does the model reproduce measured spread rates in experimental burns, documented perimeters on real incidents, or a laboratory fire in a wind tunnel? It requires data from the world and always leaves a residual, because the observations themselves have error. The rule for reading any claim about a model is to ask which of the two questions the offered evidence answers. A model can be beautifully verified and physically wrong; a model can match one famous fire and still contain a coding error that shows up at another scale. The confusion resolved is treating the word validated as a general seal of approval. After this Concept you can classify a piece of evidence about a fire model as verification or validation and name what it leaves unanswered.
Verification and validation are two distinct, complementary processes for assessing the predictive capability of deterministic computational models (per ASTM E1355): verification is the process of de…